Wildlife hosts predict the distribution of reported coccidioidomycosis in the western United States

Global environmental change is reshaping human exposure to zoonotic and environmentally acquired pathogens, yet predicting disease risk remains challenging. High-resolution risk maps typically rely on human case data and environmental correlates, often overlooking ecological processes such as wildlife reservoirs. We evaluated whether mammalian reservoir distributions improve prediction of coccidioidomycosis (Valley fever), an emerging, environmentally-acquired fungal disease with a poorly characterized range. Using county-level coccidioidomycosis notification data from the Centers for Disease Control and Prevention, we developed a hierarchical Bayesian model of county-level endemicity, defined as >=10 cases per 100,000 population. We incorporated climatic, environmental, and vegetation covariates, a state-level reporting effect, and species distribution models for 22 mammalian species previously identified as Coccidioides reservoirs. We found that the number of endemic mammalian reservoirs in a county was the strongest predictor of coccidioidomycosis endemicity, with each standard deviation increase in reservoir species richness associated with substantially higher odds of endemicity (log-odds ratio = 1.702; 95% CI: 1.060-2.419). In contrast, maximum vapor pressure deficit, soil moisture, and land cover were not independently associated with endemicity after accounting for reservoir distributions. State-level reporting effects revealed substantial heterogeneity, and comparison of models with and without reporting effects identified regions likely to be endemic but underreported, including parts of Nevada, Utah, New Mexico, Texas, and Colorado. Our results establish reservoir diversity as a central predictor of zoonotic fungal disease risk and demonstrate a transferable framework for distinguishing between ecological drivers of infection from surveillance bias to improve disease risk mapping and identify areas of potential underreporting.